Lightweight Real-Time Image Semantic Segmentation Network Based on Multi-Resolution Hybrid Attention Mechanism

Table 4

To verify the impact of each module on the final accuracy of our network, we validate the mIoU results on the Cityscapes set.

Method	Backbone	Up-sampling	Aux loss	AFM	mIoU
FCN-Res34	Resnet34	×	×	×	69.50
MHANet-Base	D-Resnet	SF	×	×	71.23(+1.73)
MHANet-Base/loss	D-Resnet	SF	√	×	72.04(+2.54)
MHANet	D-Resnet	SF	√	√	73.16(+3.66)